solve milliseconds error

Former-commit-id: 3820bdd8e592b1154620f906a602ea0d7a3a5373

SRT.py

@@ -8,8 +8,18 @@ class SRT_segment(object):
     def __init__(self, *args) -> None:
         if isinstance(args[0], dict):
             segment = args[0]
-            self.start_time_str = str(0)+str(timedelta(seconds=int(segment['start'])))+',000'
-            self.end_time_str = str(0)+str(timedelta(seconds=int(segment['end'])))+',000'
+            start_ms = int((segment['start']*100)%100*10)
+            end_ms = int((segment['end']*100)%100*10)
+            start_time = str(timedelta(seconds=int(segment['start']), milliseconds=start_ms))
+            end_time = str(timedelta(seconds=int(segment['end']), milliseconds=end_ms))
+            if start_ms == 0:
+                self.start_time_str = str(0)+start_time.split('.')[0]+',000'
+            else:
+                self.start_time_str = str(0)+start_time.split('.')[0]+','+start_time.split('.')[1][:3]
+            if end_ms == 0:
+                self.end_time_str = str(0)+end_time.split('.')[0]+',000'
+            else:
+                self.end_time_str = str(0)+end_time.split('.')[0]+','+end_time.split('.')[1][:3]
             self.source_text = segment['text']
             self.duration = f"{self.start_time_str} --> {self.end_time_str}"
             self.translation = ""
@@ -66,7 +76,7 @@ class SRT_script():
         return final_seg
 
     def form_whole_sentence(self):
-        merge_list = [] # a list of indices that should be merged e.g. [[0], [2, 3, 4], [5, 6], [7]]
+        merge_list = [] # a list of indices that should be merged e.g. [[0], [1, 2, 3, 4], [5, 6], [7]]
         sentence = []
         for i, seg in enumerate(self.segments):
             if seg.source_text[-1] == '.':
@@ -100,6 +110,8 @@ class SRT_script():
             if i < len(lines):
                 if "(Note:" in lines[i]: # to avoid note
                     lines.remove(lines[i])
+                    if i == len(lines) - 1:
+                        break
                 seg.translation = lines[i].split("：")[1]
         pass
     

pipeline.py

@@ -89,7 +89,6 @@ srt_file_en = args.srt_file
 
 if srt_file_en is not None: 
     srt = SRT_script.parse_from_srt_file(srt_file_en)
-    script_input = srt.get_source_only()
 else:
     # using whisper to perform speech-to-text and save it in <video name>_en.txt under RESULT PATH.
     srt_file_en = "{}/{}/{}_en.srt".format(RESULT_PATH, VIDEO_NAME, VIDEO_NAME)
@@ -115,9 +114,7 @@ else:
         # transcript.to_srt_vtt(srt_file_en)
         transcript = transcript.to_dict()
         srt = SRT_script(transcript['segments']) # read segments to SRT class
-        srt.form_whole_sentence()
-        script_input = srt.get_source_only()
-        srt.write_srt_file_src(srt_file_en)
+
         #Write SRT file
         
         # from whisper.utils import WriteSRT
@@ -126,55 +123,18 @@ else:
         #     writer.write_result(transcript, f)
     else:
         srt = SRT_script.parse_from_srt_file(srt_file_en)
-        script_input = srt.get_source_only()
+
+# srt preprocess
+srt.form_whole_sentence()
+srt.correct_with_force_term()
+srt.write_srt_file_src(srt_file_en)
+script_input = srt.get_source_only()
 
 if not args.only_srt:
     from srt2ass import srt2ass
     assSub_en = srt2ass(srt_file_en, "default", "No", "Modest")
     print('ASS subtitle saved as: ' + assSub_en)
-
-# # force translate the starcraft2 term into chinese according to the dict
-# # TODO: shortcut translation i.e. VA, ob
-# # TODO: variety of translation
-# from csv import reader
-# import re
-
-# # read dict
-# with open("finetune_data/dict.csv",'r', encoding='utf-8') as f:
-#   csv_reader = reader(f)
-#   term_dict = {rows[0]:rows[1] for rows in csv_reader}
-
-# def clean_timestamp(lines):
-#   new_lines = []
-#   strinfo = re.compile('[0-9]+\n.{25},[0-9]{3}')    # 注意用4个\\\\来替换\
-#   new_lines = strinfo.sub('_-_', lines)
-#   print(new_lines)
-#   return new_lines
-
-
-# ready_lines = re.sub('\n', '\n ', script_input)
-# ready_words = ready_lines.split(" ")
-# i = 0
-# while i < len(ready_words):
-#   word = ready_words[i]
-#   if word[-2:] == ".\n" :
-#     if word[:-2].lower() in term_dict :
-#       new_word = word.replace(word[:-2], term_dict.get(word[:-2].lower())) + ' '
-#       ready_words[i] = new_word
-#     else :
-#       word += ' '
-#       ready_words[i] = word
-#   elif word.lower() in term_dict :
-#       new_word = word.replace(word,term_dict.get(word.lower())) + ' '
-#       ready_words[i] = new_word
-#   else :
-#     word += " "
-#     ready_words[i]= word
-#   i += 1
-
-# script_input_withForceTerm = re.sub('\n ', '\n', "".join(ready_words))
-
-# srt.correct_with_force_term()
+    
 
 # Split the video script by sentences and create chunks within the token limit
 def script_split(script_in, chunk_size = 1000):